Skip to contents

Custimized region search for odd Weibull distribution

Source: R/myOW_region.R
myOW_region.Rd

This function can be used to modify OW gamlss.family object in order to set a customized region search for gamlss() function.

Usage

myOW_region(family = OW, valid.values = "auto", initVal)

Arguments

family

The OW family. This arguments allows the user to modify input arguments of the family, like the link functions.

valid.values

a list of character elements specifying the region for sigma and/or nu. See Details and Examples section to learn about its use.

initVal

An initValOW object generated with initValuesOW function.

Value

Returns a gamlss.family object which can be used to fit an OW distribution in the gamlss() function.

Details

This function was created to help users to fit OW distribution easily bounding the parametric space for sigma and nu.

The valid.values must be defined as a list of characters containing a call of the all function.

Author

Jaime Mosquera Gutiérrez jmosquerag@unal.edu.co

Examples

# Example 1
# Generating some random values with
# known mu, sigma and nu
y <- rOW(n=200, mu=0.2, sigma=4, nu=0.05)

# Custom search region
myvalues <- list(sigma="all(sigma > 1)",
                 nu="all(nu < 1) & all(nu < 1)")

my_initial_guess <- initValuesOW(formula=y~1)
summary(my_initial_guess)
#> --------------------------------------------------------------------
#> Initial Values
#> sigma = 5
#> nu = 0.1
#> --------------------------------------------------------------------
#> Search Regions
#> For sigma: all(sigma > 1)
#> For nu: all(nu < 1/sigma)
#> --------------------------------------------------------------------
#> Hazard shape: Bathtub

# OW family modified with 'myOW_region'
require(gamlss)
myOW <- myOW_region(valid.values=myvalues, initVal=my_initial_guess)
mod1 <- gamlss(y~1, sigma.fo=~1, nu.fo=~1, 
               sigma.start=param.startOW('sigma', my_initial_guess), 
               nu.start=param.startOW('nu', my_initial_guess),
               control=gamlss.control(n.cyc=300, trace=FALSE),
               family=myOW)

exp(coef(mod1, what='mu'))
#> (Intercept) 
#>   0.1794033 
exp(coef(mod1, what='sigma'))
#> (Intercept) 
#>    4.597043 
exp(coef(mod1, what='nu'))
#> (Intercept) 
#>  0.04445138 

# Example 2
# Same example using another link function and using 'myOW_region'
# in the argument 'family'
mod2 <- gamlss(y~1, sigma.fo=~1, nu.fo=~1, 
               sigma.start=2, nu.start=0.1,
               control=gamlss.control(n.cyc=300, trace=FALSE),
               family=myOW_region(family=OW(sigma.link='identity'),
                                  valid.values=myvalues,
                                  initVal=my_initial_guess))

exp(coef(mod2, what='mu'))
#> (Intercept) 
#>   0.1794558 
coef(mod2, what='sigma')
#> (Intercept) 
#>    4.595219 
exp(coef(mod2, what='nu'))
#> (Intercept) 
#>  0.04446516